Dynamic AI Ultrasound-Assisted Diagnosis System to Reduce Unnecessary Fine Needle Aspiration of Thyroid Nodules.
Authors
Affiliations (3)
Affiliations (3)
- Department of Ultrasound, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China.
- Department of Breast Surgery, Wujin Hospital Affiliated With Jiangsu University, Wujin, China.
- Department of Ultrasound, Xuzhou Centeral Hospital of Bengbu Medical College, Xuzhou, China.
Abstract
This study aims to compare the diagnostic efficiency of the American College of Radiology-Thyroid Imaging, Reporting, and Data System (ACR-TIRADS), fine-needle aspiration (FNA) cytopathology alone, and the dynamic artificial intelligence (AI) diagnostic system. A total of 1035 patients from three hospitals were included in the study. Of these, 590 were from the retrospective dataset and 445 cases were from the prospective dataset. The diagnostic accuracy of the dynamic AI system in the thyroid nodules was evaluated in comparison to the gold standard of postoperative pathology. The sensitivity, specificity, ROC, and diagnostic differences in the κ-factor relative to the gold standard were analyzed for the AI system and the FNA. The dynamic AI diagnostic system showed good diagnostic stability in different ages and sexes and nodules of different sizes. The diagnostic AUC of the dynamic AI system showed a significant improvement from 0.89 to 0.93 compared to ACR TI-RADS. Compared to that of FNA cytopathology, the diagnostic efficacy of the dynamic AI system was found to be no statistical difference in both the retrospective cohort and the prospective cohort. The dynamic AI diagnostic system enhances the accuracy of ACR TI-RADS-based diagnoses and has the potential to replace biopsies, thus reducing the necessity for invasive procedures in patients.